A memory die typically has one or more data latches used in the process of writing data to and reading data from non-volatile memory in the memory die. In memory systems with multiple dies, the use of the data latches allows one die to sense data and store it in a data latch for a later transfer, while data is being transferred out of another memory die. However, during mixed traffic loads where there are both long sequential reads/writes and short fragment reads/writes, memory dies performing the short fragment reads/writes need to be idle until the long sequential reads/writes are completed. To increase memory die efficiency, the memory dies performing the short fragment reads/writes can be provided with more read/write requests to keep them busy while they are waiting for other memory dies to finish and release the data bus going to the controller. However, many current architectures limit the number of read commands to one. Further, the queue depth to hold data from those additional commands is limited to the number of data latches in the memory die.